1
Empowering Providers with Data to Affect
Behavior and Reduce Cost
Session #297, February 15, 2019
Lauren Onken, MHA , Executive Director of Heart & Vascular, UNC Health Care
Shaun McDonald, Enterprise Architect – Analytics, UNC Health Care
2
Lauren Onken, MHA
Has no real or apparent conflicts of interest to report.
Shaun McDonald
Has no real or apparent conflicts of interest to report.
Conflict of Interest
3
• UNC Health Care System Overview
• Local Problem
• Design and Implementation
• Leveraging Health IT
• Value Derived
• Next Steps
Agenda
4
• Recognize the value of detailed data in driving change and
program development
• Illustrate how to statistically account for variations and complexity
• Demonstrate how to affect behaviors through the use of accurate
and meaningful data
Learning Objectives
5
Integrated, not-for-profit health care system, owned by the
State of North Carolina and based in Chapel Hill. We provide
comprehensive patient care, facilitate physician education and
research excellence, and promote the health and well-being of all
North Carolinians.
UNC Health Care System Overview
Key Stats
2011 2017
Net patient
revenues
$2.0B $4.9B
Licensed
beds 1,530 >3,400
Employees
14,000 >31,500
Medica
l staff 3,186 >5,400
Employed MDs
2,110 >3,200
Surgeries
60,000 >120,000
ED visits
151,000 >510,000
Clinic visits
1.1M >3.5M
6
Providing high quality care across the state
7
Pharmaceutical costs are rising and impacting
patient care
Percent
23
Percent
90
Percent
Increase in annual inpatient drug spending from FY13-FY15
(5.2 to 6.5M)
Of hospitals reported that inpatient drug prices increases had a
moderate or severe effect on their ability to manage costs
• Rising pharmaceutical costs threaten patient access to drug
therapies, but also challenge providers’ ability to deliver value-
based care to their patients
• The American Hospital Association (AHA) and the Federation of
American Hospitals (FAH) commissioned a study at the University
of Chicago in 2016 to better understand how drug prices are
changing in the inpatient hospital setting [1]
8
• Patient: 57 year old male presents to ER with crushing
chest pains
• Exam: Diaphoretic and restless, blood pressure 90/69, heart
rate 110
• EKG: ST elevation anterior leads
• Diagnosis:STEMI (acute heart attack)
Case Study: Percutaneous Coronary
Intervention (PCI)
9
• 2008: “HORIZONS-AMI” study Resulted in a nationwide
increase in bivalirudin use for all PCI
• 2008 – 2014:
• Transradial PCI becomes more widespread reducing bleeding risk
• Sporadic cases of stent thrombosis with bivalirudin reported in the literature
• 2014: “HEAT-PPCI” study challenged the use of bivalirudin
• Suggested that a heparin strategy reduced the incidence of major adverse
ischaemic events with no increase in bleeding complications
• Showed bivalirudin was about 300 times more expensive than heparin. It was
estimated that switching to heparin would reduce the cost of their annual 1000
PPCI cases by £500 000, ~ $640,000 (US dollars)
• Despite the new information from the “HEAT-PPCI” study and our own
cost data, many physicians were reluctant to stop using the bivalirudin.
Best practices changed with new studies
Sources : [2] NEJM. Bivalirudin during Primary PCI in Acute Myocardial Infarction (2008)
[3] American College of Cardiology. How Effective are Antithrombotic Therapies in Primary PCI – HEAT PCI (2014)
10
• Baseline: Percentage of PCI cases using bivalirudin was
88.24% at UNC Medical Center and 77.47% at UNC REX
• High pharmaceutical costs + high utilization = high
procedure costs and questionable value
Bivalirudin was the primary anticoagulant
used for PCIs
0
10
20
30
40
50
60
70
80
90
100
2014-01 2014-02 2014-03 2014-04 2014-05 2014-06
% of PCI Cases Using Bivalirudin
UNC MC
UNC REX
11
Our journey, from a heparin-first pilot to the
creation of a dashboard to drive change
UNCMC begins
piloting a heparin-first
strategy during PCI
procedures
7/2014
2/2015
UNCMC determines that a
heparin-first approach is
safe in labs with significant
rates of radial access*
12/2014
Preliminary
results shared
with UNC REX
5/2015
Centralizing PCI
reporting
incorporating external
NCDR data into a
dashboard begins
6/2015
“PCI Value Strategy”
is shared with
executive leadership
for approval
12/2015
A review of cost savings
opportunities (“Carolina Value”)
across the System
independently confirms this
cost savings opportunity
4/2016
Dashboard demoed
and PCI procedure
results and bivalirudin
usage, are shared with
providers
12
A committee was formed with multiple subject
matter experts (SMEs)
Governance Committee
Structure and Focus
Core Multidisciplinary committee (e.g.
physician and operational leaders,
analysts, etc.)
Initial work identified opportunities to:
1) Standardize best practices
2) Improve quality
3) Reduce direct costs, improve current
contribution margin, and position
UNCHCS for bundle payment success
4) Improve operational efficiency and
throughput
Selected opportunities included:
1) Reduce use of bivalirudin & substitute
heparin
2) Utilize lower cost routine supplies
3) Reduce variation in supply usage
13
This project was well-suited for the DMAIC
methodology
Define
Measure
Analyze
Improve
Control
• Cath Lab leadership identified gaps
• KPIs selected to show improvement in focus areas (i.e. “PCI Bivalirudin Usage”)
• Brought together experts from each Cath Lab for measure development
• Dashboards developed to monitor key metrics
• Compared data to source system for accuracy
• Bivalirudin identified as opportunity for cost reduction & standardization
• Validated internally derived measures against data submitted to national registries
(NCDR)
• Used dashboard drill-down functionality to identify providers who have not yet adopted
recommendation
• Showed providers their own and peer comparisons
• Discussed utilization, costs, and benefits occur with individual providers
• Dashboard continually monitored and leveraged for improvement efforts
• Leadership reviews monthly at executive steering committees
14
To create our PCI dashboard, we used our
standard development process/approach
Goals Requirements Build Test Distribute/Share
• Create a system-
wide solution
• Centralize key
metrics pulled
from multiple
systems
• Create a shared
understanding of
performance and
opportunities
• Near real-time
data so progress
and initiatives can
be regularly
monitored
• Opportunities
should be easily
identifiable and
visible
• Source systems
identified, data
mapped to data
model, and ETL to
move the data into
our data warehouse
• Reporting layer
created within BI
tool to simplify
reporting efforts
• Metric logic built
into objects to pre-
calculate values
• Results are
validated against
data submissions
to national
registries
• Source system
validation to
ensure data
matches what is
shown
• Metric calculation
documentation
created
• Made dashboard
accessible to
executive and
analysts
supporting these
departments
• Providers review
their own data to
understand
performance and
opportunities
Iterate
15
To gain a comprehensive view of our PCI events, we married
data elements stored across multiple databases
After gathering initial requirements, we
married disparate data sources
Integrating Data Into the Data Warehouse
1. Carolina Data Warehouse for Health
• Enterprise Data Warehouse contains
a data mart with cardiac data curated
for analytics
2. Lawson
• Item number, manufacturer, last
purchase price
3. RxWorks
• NDC code, lot number, last purchase
price
4. Apollo and Merge
• Supply usage/waste, event
timestamps, interventionalists, patient
encounter information
Data Elements
16
Next we created a few mock-ups, eventually
landing on our final dashboard design
Actual PCI DashboardOriginal Dashboard Mock-Ups
17
Primary Goal: Understand and share the ordering practice
patterns and behaviors to help shift the mindset of our
providers
The goal of this dashboard was to “change
behavior”
Educate providers about the cost of drugs
Demonstrate maintained or improved outcomes as a
result of the switch from bivalirudin to heparin
Minimize case-complexity and population
differences by using statistical clustering
Educate providers about their peer’s usage
18
Using the PCI dashboard - Functionality
Flexible Controls
Side-by-Side Comparisons
Standardized Data
Filters Currently Applied
Highlight Metrics
19
Providing accurate and meaningful data can
change physician behavior
Sharing Data With Colleagues
“The cardiologists received the information in an extremely
positive fashion. After training, most doctors and us
interventional cardiologists work independently during
cases and patient care with limited exposure to our
partners. Being able to share practice patterns and
understand cost/quality among peers resulted in positive
change in behavior.” - Dr. Joel Schneider, UNC REX
Physician Champion
• The dashboard was shared at many meetings and other forums to
maximize awareness
• Physicians received un-blinded dashboards and had access to their
colleagues’ information. Staff received blinded dashboards.
20
Significant reduction in the percentage of PCI
cases using Bivalirudin
Bivalirudin usage declined significantly during this initiative and
has consistently remained very low
0
20
40
60
80
100
REX UNCMC
% of PCI Cases Using Bivalirudin
Pre
Post
There’s been a 90% percent reduction in Bivalirudin
usage since the start of this effort
21
The dashboard played a key role in supporting
education and adoption efforts
0
10
20
30
40
50
60
70
80
90
100
Provider 1 Provider 2 Provider 3
Provider 4 Provider 5 Provider 6
Provider 7 Provider 8
Visible change in behavior after education and dashboard roll-out
Provider Education
Starts April 2016
% of PCI Cases Using Bivalirudin
22
This reduction represents a total savings of
~$1.6 million annually
• Savings is calculated as the cost difference between Heparin
and Bivalirudin per administration multiplied by the number of
cases
• ROI – Savings for CY17
Item
Amount
Additional Hardware Infrastructure Required
+ $0
Heparin Medication Cost Per Administration + $4.68
Bivalirudin Medication Savings Per Administration - $727.65
Total Savings from Education in CY17
(284 cases during CY17 – 8 providers)
- $176K
Total Savings from overall initiative in CY17
(2,886 cases during CY17)
- $ 1.6
Million
23
Most importantly, this substitution has not
impacted patient quality or safety
Adverse events and vascular complications are the primary
measures monitored to ensure that anticoagulation medications are
working correctly and preventing harm
In-Stent Thrombosis (%)
0
5
10
15
20
25
2013-Q4
2014-Q1
2014-Q2
2014-Q3
2014-Q4
2015-Q1
2015-Q2
2015-Q3
2015-Q4
2016-Q1
2016-Q2
2016-Q3
2016-Q4
2017-Q1
2017-Q2
2017-Q3
REX UNCMC
In-Stent Thrombosis %
24
Most importantly, this substitution has not
impacted patient quality or safety
Adverse events and vascular complications are the primary
measures monitored to ensure that anticoagulation
medications are working correctly and preventing harm
Observed vs Expected Mortality Ratio
0
2
4
6
8
10
12
14
UNC REX UNC MEDICAL CENTER
Observed/Expected Mortality
25
UNC HC took an education-first approach to
change influence provider behavior
• Overwhelming physicians with additional, disruptive prompts
could increase alert fatigue and lead to patient safety events
[1]
• Intrusive provider alerts are only used as a means of last resort or
when the prevention of adverse events warrants such an interruption
• Educating providers first is the most successful method to
changing behavior
• Provider level un-blinded data works best when providers know that
leadership is also reviewing the same data and has been effective with
the following:
• Education on process or performance
• Best practice sharing
• Healthy competition
Sources : [1] Patient Safety Primer – Alert Fatigue, https://psnet.ahrq.gov/primers/primer/28/alert-fatigue, – Agency for Healthcare Research and Quality (2018)
26
Next Steps – Continuous Improvement
1. The PCI dashboard allows continued monitoring for adverse
events as well as other supply cost opportunities for savings
and standardization
2. The translational model of moving positive findings from one
institution to another by leveraging comparative can lead to
other value-added opportunities at other hospitals across
the Health Care System
3. Analyzing and sharing practice patterns through the
dashboard can identify new opportunities and influence
individual physicians to adopt best practices
4. Scale this concept to other procedures and entities in
cardiology such as AICD implantation, STEMI management,
stress testing, etc.
